Small, modified and highly structured: the challenge of tRNA sequencing

Irina Chelysheva (University of Hamburg) began by explaining how, when her lab first received a MinION, they were keen to use it to "try something completely new": they settled on the sequencing of tRNAs. She introduced the challenge that this posed: tRNAs are short, at 76-90 nucleotides, structurally stable and are the most highly modified RNA species, with at least 40 modifications present per tRNA. Irina described the benefits of using the MinION: "it's portable, it's quick, it's cheap" and modifications can be studied "very well."

Irina introduced how she and her colleague Priyanka Nair developed a method for preparing and sequencing tRNAs on the MinION, with Priyanka optimising the processes of ligation and clean-up of the small molecules. They first decided to use the Direct RNA Sequencing Kit (without reverse transcription), with a custom-designed adapter featuring sequence complementary to the 3' CCA tail of tRNAs. For their first run, they chose to sequence in vitro transcribed standard tRNA molecules, as these do not contain any modifications, they were able to define the concentration of their input and they had known reference sequences. Choosing proline tRNA from E.coli, the sample was prepared and sequenced on the MinION and basecalled. However, when it came to alignment, a global alignment approach resulted in <1% aligned reads, with 97% for the control RNA sample; they concluded that global alignment was not suitable due to the short reads. Local alignment was chosen, in which several regions are aligned rather than as much as possible, though there can be gaps: this resulted in 44% of reads aligning.

Next, they decided to try sequencing four tRNAs pooled to equimolar concentrations. Using the local alignment approach, 88.6% of the reads aligned to one or more references, with 32% uniquely classified as one type of tRNA.  However, Irina showed how one tRNA in particular was poorly represented; she suggested that not all samples were ligating to adapters with the same efficiency. She noted that this demonstrated that it was possible to sequence tRNAs on a MinION and faithfully classify a reasonable proportion as particular tRNA species, but that the method was not yet quantitative. To address the poorer coverage of the 3' end of the tRNAs, important in the sequencing of especially short RNAs, they decided to add a 3' polyA tail to their tRNAs. This both added sequence to the 3' end, extending their length slightly, and enabled the use of the original polyT adapters from the Direct RNA Sequencing Kit. Showing the results for the same mix of four tRNAs, this time using the polyA method, Irina pointed out that whilst the quality didn't significantly improve, the clusters for the four tRNAs were better defined and the alignment score increased, with longer aligned sections produced. The same tRNA as in the previous test was again poorly represented, suggesting the issue for that species was ligation-based.

Following on from this, they decided to use this protocol in the sequencing of total native tRNAs. However, this time, the tRNAs failed to align and cluster. To explain this result, Irina highlighted the many modifications present in the short tRNA sequences, which could have lead to miscalls. She said she was excited to test the Taiyaki basecaller to account for these modifications.

In summary, Irina concluded that "overall, nanopore sequencing is applicable to tRNAs"; after only six months of working with nanopore technology, she said that she hoped to show more data from the project next year, and was looking forward to discussing the application with other delegates to hear their ideas.